Comparison semantics · version 1

Comparability is computed, not assumed.

You may inspect any runs together. Runpile classifies material differences against your intent so a convenient table never becomes a misleading ranking.

incompatible

Direct ranking is misleading

Examples: warm shared-prefix cache versus cold unique prompts; generation versus embedding.

major

Useful only with a prominent caveat

Examples: fixed 6,000/1 versus 6,144/1; synthetic versus real data; different timing boundaries.

minor

Likely comparable with a caveat

Examples: a runtime patch-level change where workload and metric definitions match.

informational

An intentional comparison dimension

Examples: RTX 5090 versus H100 when the selected intent is comparing hardware.